Performance of Different Risk Scores for the Detection of Atrial Fibrillation Among Patients With Cryptogenic Stroke.

Background: Atrial fibrillation (AF) is a frequent underlying cause of cryptogenic stroke (CS) and its detection can be increased using implantable cardiac monitoring (ICM). We sought to evaluate different risk scores and assess their diagnostic ability in identifying patients with CS with underlying AF on ICM.

Methods: Patients with CS, being admitted to a single tertiary stroke center between 2017 and 2022 and receiving ICM, were prospectively evaluated. The CHADS-VASc, HAVOC, Brown ESUS-AF, and CHEST scores were calculated at baseline. The primary outcome of interest was the detection of AF, which was defined as at least 1 AF episode on ICM lasting for 2 consecutive minutes or more. The diagnostic accuracy measures and the net reclassification improvement were calculated for the 4 risk scores. Stroke recurrence was evaluated as a secondary outcome.

Results: A total of 250 patients with CS were included, and AF was detected by ICM in 20.4% (n=51) during a median monitoring period of 16 months. Patients with CS with AF detection were older compared with the rest (=0.045). The median HAVOC, Brown ESUS-AF, and CHEST scores were higher among the patients with AF compared with the patients without AF (all <0.05), while the median CHADS-VASc score was similar between the 2 groups. The corresponding C statistics for CHADS-VASc, HAVOC, Brown ESUS-AF, and CHEST for AF prediction were 0.576 (95% CI, 0.482-0.670), 0.612 (95% CI, 0.523-0.700), 0.666 (95% CI, 0.587-0.746), and 0.770 (95% CI, 0.699-0.839). The CHEST score presented the highest diagnostic performance based on C statistics (<0.05 after correction for multiple comparisons) and provided significant improvement in net reclassification for AF detection (>70%) compared with the other risk scores. Finally, stroke recurrence was documented in 5.6% of the study population, with no difference regarding the 4 risk scores between patients with and without recurrent stroke.

Conclusions: The CHEST score was superior to the CHADS-VASc, HAVOC, and Brown ESUS-AF scores for discriminating patients with CS with underlying AF using ICM.